KR20180131381A - Stand-along Voice Recognition based Agent Module for Precise Motion Control of Robot and Autonomous Vehicles - Google Patents

Abstract

An objective of the present invention is to provide a stand-alone embedded voice recognition technique capable of precisely recognizing and analyzing a robot control voice command even in a space without internet connection and a voice recognition-based agent module capable of precise motion control of an autonomous vehicle. According to embodiments of the present invention, the stand-alone voice recognition-based agent module for precise motion control of a robot and an autonomous vehicle comprises: a voice recognition engine to perform language processing to output an inputted voice in recognizable text and then convert the recognizable text into speed control text for motion control to assign the speed control text; a robot operating system to receive a speed control command using a programming language, to which the speed control text is assigned by the voice recognition engine, to transfer a speed value to drive a robot or an autonomous vehicle; and a driving unit to receive the speed value from the robot operating system to drive the robot or the autonomous vehicle.

Description

[0001] The present invention relates to a stand-alone voice recognition based agent module for precise motion control of a robot and an autonomous mobile body,

The present invention relates to a standalone speech recognition based agent module for robotic and autonomous mobile body precise motion control. More particularly, the present invention relates to a voice recognition based on a voice recognition algorithm and a voice recognition algorithm including an acoustic model, a language model, The present invention relates to a robot capable of precisely controlling motion of a robot or an autonomous mobile body, and a standalone voice recognition based agent module for precise motion control of an autonomous mobile body.

Generally speaking, speech recognition refers to a technique of transferring voice to a word or sentence through a wired / wireless communication method such as a microphone. Such speech recognition can be added to human convenience and has been widely used in the fields of robotics, mobile phones, smart homes, automobiles, and infotainment.

As a result, speech recognition has been studied steadily, and active research is expected to continue in the future.

Recently, Open Source module for voice recognition has been disclosed, and smart home based on voice recognition using an open module such as Amazon Alexa, Google Home, etc. and personal secretary have been attracting attention.

However, in the case of such an open source module, a stand-alone embedded speech recognition module capable of freely and precisely controlling the robot even in a space without an internet connection is provided because the service is provided only through an Internet connection Development is required.

In order to solve the above problems, the present invention has been made to solve the above-mentioned problems, and it is an object of the present invention to provide a speech recognition system capable of precisely recognizing and interpreting a robot- It is an object to provide an agent module.

In order to accomplish the above object, there is provided a standalone speech recognition based agent module for robotic and autonomous moving object precise motion control according to an embodiment of the present invention, language processing for outputting an input voice as recognizable text, A speech recognition engine for converting and assigning speed control text; A robot operating system that receives a speed control command through a programming language that is assigned a speed control text in the speech recognition engine and transmits a speed value such that the robot or the autonomous mobile is driven; And a driving unit that receives the velocity value from the robot operating system and drives the robot or the autonomous mobile unit.

Here, the speech recognition engine may include Pocketphinx or other open speech recognition module.

The speech recognition engine may include a preprocessing unit for extracting a feature vector necessary for speech recognition and a recognition unit for storing the speech recognition algorithm and analyzing the feature vector extracted by the preprocessing unit through the speech recognition algorithm and performing language processing can do.

The speech recognition algorithm may include an acoustic model for adapting speech input through a microphone; A language model for converting the extracted feature vector into an recognizable text form by comparing the extracted feature vector with the adaptive acoustic model and a language model for converting the extracted feature vector into a recognizable text form when the extracted feature vector and the acoustic model are compared Quot; data dictionary "

Also, the programming language may be Python or C / C ++.

In addition, the speed control command transfer from the programming language to the robot operating system may be carried out via a wired or wireless communication scheme.

The driving unit may include a low-level processor for receiving a velocity value from the robot operating system and inputting the velocity as a low-level signal; A PWM generator for pulse-modulating the speed signal input through the low-level processor, and a DC motor for driving the robot or the autonomous mobile body according to the pulse modulated from the PWM generator.

The robot and the independent speech recognition based agent module for precise motion control of the autonomous moving object according to the embodiment of the present invention are characterized in that the robot can be freely and precisely controlled even in a space without internet connection.

In addition, more accurate speech recognition can be performed by using the speaker adaptation technique of Maximum Likelihood Linear Regression (MLLR) and Maximum A Posteriori (MAP) as an acoustic model of the present invention.

In addition, by using an open speech recognition engine including an open speech recognition engine Pocketsphinx, it can be provided at a relatively low price.

1 is a block diagram illustrating a configuration of a standalone speech recognition based agent module for robotic and autonomous moving object precision motion control according to an embodiment of the present invention.
2 is a block diagram showing a configuration of a speech recognition engine which is an embodiment of the present invention.
3 is a block diagram showing the configuration of a recognition unit which is a configuration of the speech recognition engine of FIG.
4 is a block diagram showing a speaker adaptation step of an acoustic model which is an embodiment of the present invention.
5 is an operational flowchart of a standalone speech recognition based agent module for robotic and autonomous moving object precision motion control according to an embodiment of the present invention.
FIG. 6 is an operational flowchart of the step (c) of FIG.
7 is an operational flowchart of the step (f) of Fig.

Hereinafter, the description of the present invention with reference to the drawings is not limited to a specific embodiment, and various transformations can be applied and various embodiments can be made. It is to be understood that the following description covers all changes, equivalents, and alternatives falling within the spirit and scope of the present invention.

In the following description, the terms first, second, and the like are used to describe various components and are not limited to their own meaning, and are used only for the purpose of distinguishing one component from another component.

Like reference numerals used throughout the specification denote like elements.

As used herein, the singular forms "a", "an" and "the" include plural referents unless the context clearly dictates otherwise. It is also to be understood that the terms " comprising, "" comprising, "or" having ", and the like are intended to designate the presence of stated features, integers, And should not be construed to preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, or combinations thereof.

Hereinafter, embodiments of the present invention will be described in detail with reference to FIGS. 1 to 7.

FIG. 1 is a block diagram showing a configuration of a standalone speech recognition based agent module for robotic and autonomous moving object precise motion control according to an embodiment of the present invention. FIG. 2 is a block diagram illustrating a configuration of a speech recognition engine, FIG. 3 is a block diagram showing a configuration of a recognition unit which is a constitution of the speech recognition engine of FIG. 2, and FIG. 4 is a block diagram showing a speaker adaptation step of an acoustic model, which is a constitution of the present invention .

1 to 4, a standalone speech recognition based agent module for robotic and autonomous moving object precise motion control according to an embodiment of the present invention is installed in a robot or an autonomous mobile body, and includes a speech recognition engine 10, A robot operating system 30, and a driving unit 40. [

Specifically, the speech recognition engine 10 can recognize a voice input from a robot or a microphone installed in an autonomous mobile unit and a wired / wireless communication-based voice transmission apparatus. In addition, the speech recognition engine 10 may be configured as a PocketSphinx.

This is an open speech recognition engine and has advantages of low cost, high compatibility with a programming language such as a robot operating system (ROS) and C / C ++ or Python configured to control a robot or an autonomous mobile, have. However, the above-described pocketshapinx is a preferred example, and the present invention is not limited thereto and may be provided with another speech recognition engine 10.

Hereinafter, a description will be made on the basis of a speech recognition engine 10 composed of a preferred example PocketSpinx.

The speech recognition engine 10 can perform language processing for outputting voice inputted from the installed microphone and the wired / wireless communication based voice delivery device as recognizable text, and converts the language processed text into speed control text for motion control Can be assigned.

To this end, the speech recognition engine 10 may include a preprocessing unit 12 and a recognition unit 14 as shown in FIG.

The preprocessing unit 12 can extract a feature vector necessary for speech recognition. That is, when speech is input from the microphones to the speech recognition engine 10, the preprocessing unit 12 can extract a feature vector that can express phonetic features from the speech. At this time, the preprocessing unit 12 can extract the feature vector in units of 1/100 (seconds).

In addition, the preprocessing unit 12 can extract the feature vector using a Mel Frequency Cepstral Coefficients (MFCC) algorithm. Here, the MFCC algorithm divides the entire input sound into a short time, and extracts the characteristics by analyzing the spectrum of the input sound. For example, the length of a predetermined section is divided into 20 to 40 ms units, and a spectrum corresponding to each unit, that is, a frequency is calculated.

Meanwhile, the preprocessing unit 12 may extract a feature vector by performing a noise process for processing noise including channel distortion and background noise caused by a microphone, a line, and the like. In order to increase the success rate of speech recognition, it is possible to use a method of extracting a feature vector and compensating it, or introducing a feature vector resistant to noise.

Accordingly, the preprocessing unit 12 extracting the feature vector can transmit the feature vector to the recognition unit 14.

The recognition unit 14 can pattern-analyze the feature vectors extracted by the preprocessing unit 12 and perform language processing. In addition, the recognition unit 14 may store the speech recognition algorithm 16 and may perform language processing through the speech recognition algorithm 16. Here, the speech recognition algorithm 16 may include an acoustic model 16a, a language model 16b, and a data dictionary 16c, as shown in FIG.

More specifically, the recognition unit 14 receives the feature vector of the speech extracted from the preprocessing unit 12 and compares the pattern with the acoustic model 16a stored in the database of the speech recognition engine 10 to obtain the recognition result . As shown in FIG. 4, the acoustic model 16a is formed to have adaptability with the voice of the speaker transmitted through the speaker to increase the recognition rate. For this purpose, the acoustic model 16a includes Maximum Likelihood Linear Regression (MLLR) MAP (Maximum A Posteriori) adaptation technique.

At this time, the acoustic model 16a may perform Maximum A Posteriori (MAP) after adaptation of Maximum Likelihood Linear Regression (MLLR), whereby the acoustic model 16a provides a sample as close as possible to the speech of the speaker, So that the user 10 can correctly recognize the voice.

On the other hand, the acoustic model 16a may be composed of Korean or English.

The language model 16b can perform language processing on the voice recognized through the acoustic model 16a. To this end, the language model 16b may include a word unit search and a sentence unit search.

The word-based search is performed including phonemes, and candidate words or candidate phonemes can be extracted through word-by-word or phoneme-by-phoneme pattern comparison with the acoustic model 16a stored in the database. At this time, the candidate word or the candidate phoneme that has undergone the above process can be processed by a sentence unit search.

The sentence unit search can determine the most suitable word or phoneme by judging whether it matches the grammar structure, the sentence context, a specific topic or the like based on the information of candidate words or candidate phonemes using a data dictionary.

For example, assuming that it is difficult to distinguish between '' and '' by '' we go to the seashore '' by unclear pronunciation, two word candidates, 'a' and ' As a result. At this time, through the analysis of the sentence structure using the data dictionary (16c) in the sentence unit search, 'a' plays a role of investigation in the sentence, but it can be recognized that there is no investigation of 'ability' and can be excluded from the candidate.

That is, the language model 16b can perform language processing to improve recognition performance by restricting the vocabulary and grammar structure. In this way, the speech recognition of the speech recognition engine 10 is executed more quickly, and the speech recognition result can be more accurate.

Here, the language model 16b is based on statistical pattern recognition, and HMM (Hidden Markov Model), which is a combination of word unit search and sentence unit search, can be used. It stores the statistical information of the patterns corresponding to the speech unit in the form of a probability model and calculates the probability that an unknown pattern in each model comes out when an unknown input pattern comes in, .

On the other hand, the language model 16b can transmit processed texts in a text format so as to control motion of a robot or autonomous mobile body during language processing.

That is, the speech feature vector extracted from the preprocessing unit 12 is compared with the acoustic model 16a adapted to the speech of the speaker to extract candidate words or candidate phonemes through the language model 16b, The candidate word or candidate phonemes are distinguished by the correct sentence unit by discriminating the most suitable word or phoneme based on the data dictionary 16c. At this time, the speech processed to be divided into sentence units is textized through the language model 16b .

The language processed text may be converted to a speed control text that can control the speed of the motor for motion control of the robot or the autonomous mobile body through the speech recognition engine 10 as described above.

The assigned speed control text may be transferred to the robot operating system 30 by replacing it with a robot control command via the programming language 20 to control motion.

At this time, the programming language 20 may be C / C ++ or Python. Also, the programming language 20 may be connected to the robot operating system 30 in a wired or wireless manner. That is, the programming language 20, which is assigned the speed control text through the speech recognition engine 10, can generate a robot control command through an algorithm and deliver it to the robot operating system 30 in a wired or wireless manner. Here, the wireless scheme may be configured by a method such as Bluetooth or Wi-Fi.

The robot operating system 30 can receive the speed control command from the programming language 20. [ In addition, the robot operating system 30 can control the driving unit 40 by transmitting a velocity value to the robot 40 or the driving unit 40 that drives the autonomous moving object. That is, the robot operating system 30 can control the driving unit 40 so that the motion of the robot or the autonomous mobile unit is controlled by adjusting the speed value according to the speed control command from the programming language 20. [

The driving unit 40 receives the velocity value from the robot operating system 30 and can drive the robot or the autonomous mobile unit. To this end, the driving unit 40 may include a low-level processor 42, a PWM generator 44, and a DC motor 46.

Specifically, the low-level processor 42 may program the speed value transmitted through the robot operating system 30 as a low-level signal and input the low-level signal.

The PWM generator 44 can be programmed from the low-level processor 42 to a low level to modulate a pulse of the input low-level signal. That is, the velocity value can be controlled by pulse modulation of the low-level signal expressing the velocity value.

The DC motor 46 is connected to the robot or the wheels of the autonomous mobile body, and can drive the robot or the autonomous mobile body according to the pulse modulated by the PWM generator 44. At this time, it is preferable that the DC motor 46 is provided for each wheel, and the motion of the robot or the autonomous mobile body can be controlled by varying the control speed of each DC motor 46.

Hereinafter, a method for operating a standalone speech recognition-based agent module for robotic and autonomous moving object precise motion control according to an embodiment of the present invention will be described with reference to FIG. 5 to FIG.

FIG. 5 is an operational flowchart of a standalone speech recognition based agent module for robotic and autonomous moving object precise motion control according to an embodiment of the present invention, FIG. 6 is an operational flowchart of the step (c) of FIG. 5, 5 is an operation flowchart of step (f) of FIG.

5 to 7, a method of operating a standalone speech recognition based agent module for robotic and autonomous moving object precise motion control according to an embodiment of the present invention may be performed in the order of steps (a) to (f) .

(a) a user's voice input step (S100) through a microphone,

- The user can input a voice command through a microphone installed in the robot or autonomous mobile unit. At this time, the voice command can be transmitted to the voice recognition engine 10.

(b) extracting a feature vector of the input speech (S200)

When the voice command reaches the voice recognition engine 10, the preprocessing unit 12 of the voice recognition engine 10 can extract the feature vector of the input voice. Here, the speech recognition engine 10 may be provided with Pocketphinx, and the feature vector may be extracted in units of 1/100 (seconds) using an MFCC (Mel Frequency Cepstral Coefficients) algorithm.

(c) converting the feature vector into recognizable text using a speech recognition algorithm (S300)

This step may be performed in the recognition unit 14 of the speech recognition engine 10. At this time, the speech recognition algorithm 16 may be stored in the recognition unit 14. That is, the recognition unit 14 can receive the feature vector extracted from the preprocessing unit 12, and the recognition unit 14 can convert the recognition vector into the recognizable text using the stored speech recognition algorithm 16. [

Specifically, the speech recognition algorithm 16 may include an acoustic model 16a, a language model 16b, and a data dictionary 16c, and may be performed in the following three steps.

Step 1: The acoustic model 16a adapts to the speaker's voice (S310)

The acoustic model 16a can be compared with the feature vector extracted from the preprocessing unit 12 to derive the speech recognition result. In this case, considering that the phonetic characteristics are different from each speaker, the acoustic model 16a can be adapted to the speaker . Here, the acoustic model 16a may use a Maximum Likelihood Linear Regression (MLLR) and a Maximum A Posteriori (MAP) adaptation scheme. After performing MLLR adaptation, MAP (Maximum A Posteriori) adaptation is performed sequentially You can proceed.

Thus, the recognition rate can be further increased.

Step 2: The voice is recognized by comparing the acoustic model adapted to the speaker with the feature vector (S320)

If the initial acoustic model 16a adapts to the speaker in step 1, the feature vector may be compared with the adaptive acoustic model 16a. At this time, the acoustic model 16a can perform more accurate recognition through speaker adaptation.

Step 3: The language model 16b extracts the candidate phonemes or candidate words according to the recognized speech, and then the correct voice is discriminated by using the data dictionary 16c. Then, the recognition result is displayed in a recognizable text form for robot and autonomous mobile body motion control In operation S330,

- According to the voice recognized in the second stage, the language model (16b) can extract candidate phonemes or candidate words through word unit search and sentence unit search through HMM (Hidden Markov Model) technique. Here, the language model 16b can compare the extracted candidate phonemes or candidate words through a predetermined data dictionary 16c to determine the most suitable word or phoneme.

The sentence discriminated through the above process can be converted into a recognizable text form for precise motion control of the robot and the autonomous mobile.

(d) converting the text converted to be recognizable into the speed control text (S400)

The recognizable text derived through the speech recognition algorithm 16 can be converted into texts that can control the speed and delivered to the programming language 20. This step can be performed in the speech recognition engine 10.

(e) generating a speed control command through a programming language of the speed control text (S500)

The speed control text transferred to the programming language 20 can generate a speed control command through a coded algorithm. At this time, the generated speed control command may be transmitted to a robot operating system that controls the robot or the autonomous mobile body in a wired or wireless manner such as Wi-fi or Bluetooth.

Meanwhile, the programming language 20 may be provided in C / C ++ or Python.

(f) controlling (S600) a driving unit (40) for the robot operating system (30) to operate the robot or the motion of the autonomous mobile according to the speed control command,

- The robot operating system 30 receiving the speed control command can send and control a speed command (speed value) so that the driving unit 40 that operates the motion of the robot or the autonomous mobile unit sends the speed.

The driving unit 40 may include a low-level processor 42, a PWM generator 44, and a DC motor 46 to perform the following three steps.

Step S610: The low level processor 42 receives the speed value (speed command) from the robot operating system 30 and inputs the speed as a low level signal (S610)

The low-level processor 42 may receive the velocity value from the robot operating system 30. [ At this time, the low-level processor 42 receiving the velocity value can input the velocity as a low-level signal through programming.

Step S620: The PWM generator pulse-modulates the low-level signal inputted through the low-level processor 42 (S620)

- The PWM generator 44 can take the low level signal programmed through the low-level processor 42 and pulse-modulate the signal.

Step 3: The DC motor drives the robot or the motion of the autonomous mobile according to the pulse modulated by the PWM generator 44 (S630)

- The pulse modulated from the PWM generator 44 is transmitted to each DC motor 46 provided for each wheel of the robot or the autonomous mobile body so that the speed of the DC motor 46 is controlled in accordance with the pulse so that the motion of the robot and the autonomous mobile body Lt; / RTI >

Accordingly, although the conventional open type speech recognition module is provided only through an Internet connection, a standalone speech recognition based agent module for precise motion control of a robot and an autonomous moving object according to an embodiment of the present invention, and an operation method thereof, There is a feature that the robot can be freely controlled even in the space without.

While the present invention has been described in connection with what is presently considered to be practical exemplary embodiments, it is to be understood that the invention is not limited to the disclosed embodiments, but, on the contrary, You will understand. The embodiments described above are therefore to be considered in all respects as illustrative and not restrictive.

10: Speech recognition engine
12:
14:
16: Speech Recognition Algorithm
16a: Acoustic model
16b: Language model
16c: Data Dictionary
20: Programming language
30: Robot operating system
40:
42: low-level processor
44: PWM generator
46: DC motor

Claims (7)

A speech recognition engine for language processing to output the input voice as recognizable text, and then converting and assigning speed control text for motion control;
A robot operating system that receives a speed control command through a programming language that is assigned a speed control text in the speech recognition engine and transmits a speed value such that the robot or the autonomous mobile is driven;
And a driving unit for receiving the velocity value from the robot operating system and driving the robot or the autonomous mobile unit, and a standalone voice recognition based agent module for autonomous moving object precise motion control.
The method according to claim 1,
Wherein the speech recognition engine is a robot including a Pocketphinx or other open speech recognition module and a standalone speech recognition based agent module for autonomous mobile precise motion control.
The method according to claim 1,
Wherein the speech recognition engine comprises:
A preprocessor for extracting a feature vector necessary for speech recognition,
And a recognition unit for storing the speech recognition algorithm and analyzing the feature vector extracted by the preprocessing unit through the speech recognition algorithm and performing a language processing, and a standalone speech recognition based agent module for precise motion control of the autonomous mobile.
The method of claim 3,
The speech recognition algorithm includes:
An acoustic model for adapting the voice input through the microphone;
A language model that compares the extracted feature vector with the adaptive acoustic model to convert it into a recognizable text form,
A robot including a data dictionary for discriminating whether the language model can be converted into a recognizable text form when the extracted feature vector is compared with the acoustic model, and a standalone speech recognition based agent module for precision motion control of an autonomous moving object .
The method according to claim 1,
Characterized in that the programming language is Python or C / C ++, and a standalone speech recognition based agent module for precise motion control of an autonomous moving object.
The method according to claim 1,
Wherein the speed control command transmission from the programming language to the robot operating system is transmitted through a wired or wireless communication system.
The method according to claim 1,
The driving unit includes:
A low level processor for receiving a speed value from the robot operating system and inputting the speed as a low level signal;
A PWM generator for pulse-modulating a speed signal input through the low-level processor and
A robot including a DC motor for driving a robot or an autonomous mobile according to a pulse modulated from the PWM generator, and a standalone speech recognition based agent module for autonomous mobile precise motion control.

Images